Optical and Microphysical Properties of the Aerosols during a Rare Event of Biomass-Burning Mixed with Polluted Dust

Abstract

A rare event of mixed biomass-burning and polluted dust aerosols was observed over Athens, Greece (37.9° N, 23.6° E), during 21–26 May 2014. This event was studied using a synergy of a 6-wavelength elastic-Raman-depolarization lidar measurements, a CIMEL sun photometer, and in situ instrumentation. The FLEXPART dispersion model was used to identify the aerosol sources and quantify the contribution of dust and black carbon particles to the mass concentration. The identified air masses were found to originate from Kazakhstan and Saharan deserts, under a rare atmospheric pressure system. The lidar ratio (LR) values retrieved from the Raman lidar ranged within 25–89 sr (355 nm) and 35–70 sr (532 nm). The particle linear depolarization ratio (δ_aer) ranged from 7 to 28% (532 nm), indicating mixing of dust with biomass-burning particles. The aerosol optical depth (AOD) values derived from the lidar ranged from 0.09–0.43 (355 nm) to 0.07–0.25 (532 nm). An inversion algorithm was used to derive the mean aerosol microphysical properties (mean effective radius (r_eff), single scattering albedo (SSA), and mean complex refractive index (m)) inside selected atmospheric layers. We found that r_eff was 0.12–0.51 (±0.04) µm, SSA was 0.94–0.98 (±0.19) (at 532 nm), while m ranged between 1.39 (±0.05) + 0.002 (±0.001)i and 1.63 (±0.05) + 0.008 (±0.004)i. The polarization lidar photometer networking (POLIPHON) algorithm was used to estimate the vertical profile of the mass concentration for the dust and non-dust components. A mean mass concentration of 15 ± 5 μg m⁻³ and 80 ± 29 μg m⁻³ for smoke and dust was estimated for selected days, respectively. Finally, the retrieved aerosol microphysical properties were compared with column-integrated sun photometer CIMEL data with good agreement.

1. Introduction

Aerosols play a crucial role in the Earth’s radiation budget in a direct way, through scattering and absorption of the solar and terrestrial radiation, as well as by acting as cloud condensation nuclei (CCN), influencing clouds’ optical and microphysical properties (indirect way) [1,2]. Arid and semi-arid areas are contributing the most in terms of natural aerosol repositories. In the northern hemisphere, major dust source regions are located in the so-called dust belt spanning from the Saharan desert in Africa to the Taklimakan and Gobi deserts in central and East Asia [3]. Mineral dust is a significant component, representing approximately 75% of the global aerosol mass load and 25% of the global aerosol optical depth (AOD) [4]. Dust originating from different sources exhibits variations in mineralogical composition [5]; optical properties [6,7,8,9]; and, consequently, different effects on solar/Earth radiation and climate [10,11,12,13]. Typical sources of biomass-burning (BB) aerosols are forest fires (natural and anthropogenic) and agricultural and grassland burnings. BB is responsible for a larger fraction of global mean emissions of black carbon (BC) and organic carbon (OC) [14]. Additionally, BB is a significant source of carbonaceous aerosols [15]. The optical properties of smoke depend on the geographical source region, the season, and the type of BB aerosol [16,17]. Aerosols from BB sources can induce either warming or cooling effects on the climate [18,19].

Although several studies have been carried out regarding the long-range transport of polluted desert dust, mainly from the Sahara, mixed with biomass-burning aerosol from local or regional emissions [20,21,22,23,24], there are still limited cases observed of dust transport from both the Sahara and other desert regions (e.g., Syria, Aralkum, South Kazakhstan) at the same time [9,25,26,27]. These studies have mainly focused on vertical profiling of the aerosol optical and microphysical properties retrieved by multi-wavelength Raman lidars and sun-photometer data.

Motivated by these observational gaps and to further obtain representative data regarding the profiling of the aerosol optical and microphysical properties of mixed dust and biomass-burning particles, we employed a synergy of instrumentation (Raman and depolarization lidar, sun-photometer, PM Quartz filter samples, OC-EC Aerosol Analyzer) in order to obtain accurate aerosol vertical profiles of optical (backscattering and extinction coefficients, lidar ratio-LR, Ångström exponent) and microphysical (effective radius, single-scattering albedo, and complex refractive index) properties, as well as the chemical composition of dust and biomass-burning aerosols at ground level. Our data could be used to further improve our understanding of the aerosol radiative forcing mechanisms and to reduce the relevant associated uncertainties of climate-forecasting models [28,29,30,31,32,33].

Here, we present a rare case of a simultaneous transport of dust (from Kazakhstan and Saharan deserts) and biomass-burning aerosols (from South Russia), observed using remote sensing instrumentation (multi-wavelength Raman-depolarization lidar, CIMEL sun photometer), in situ aerosol sampling, and modeling (microphysical inversion schemes and FLEXPART model) to retrieve a complete set of the aerosols’ optical, microphysical, and chemical properties. This 6-day long-range transport event was observed over Athens, Greece, on 21–26 May 2014. At first, we present the instrumentation and methodology used for retrieving the aerosol properties, as well as the dust and BB particle dispersion model to justify their origin (Section 2). Our analysis regarding the properties of mixed aerosols (dust and biomass-burning), as derived from the synergy of remote sensing and in situ techniques, is presented in Section 3, with emphasis on the days with optimal lidar retrievals. Finally, in Section 4, we present our final remarks and conclusions.

2. Methodology and Instrumentation

2.1. Raman and Depolarization Lidar Systems for the Retrieval of the Aerosol Optical Properties

At the National Technical University of Athens (NTUA) a compact, 6-wavelength Raman lidar system (EOLE) was used to perform vertical profile measurements of the aerosol optical properties in the Planetary Boundary Layer (PBL) and the lower troposphere. The lidar system is based on a pulsed Nd:YAG laser emitting simultaneously at 355, 532, and 1064 nm with energies of 240, 310, and 260 mJ, respectively, at a repetition rate of 10 Hz. The elastically backscattered lidar signals (at 355, 532 and 1064 nm), as well as those produced by Raman scattering by the atmospheric N₂ (at 387 and 607 nm) and H₂O (at 407 nm), were simultaneously recorded by photomultipliers (PMTs) and avalanche photodiode systems (APD) after spectral separation of the returned lidar signals. The lidar signals detected at 355, 387, 532, 607, and 1064 nm were used to derive the aerosol backscatter (at 355, 532 and 1064 nm), the extinction (at 355 and 532 nm) coefficient, and the Ångström exponent (AE) profiles [33], while the 407 nm channel was used to derive the water vapor mixing ratio profile [34].

To achieve reliable and quantitative lidar aerosol retrievals, a combination of various techniques and methods is necessary. The elastic backscatter lidar technique proves suitable for retrieving aerosol parameters, particularly for moderate aerosol optical depths (AOD < 0.2–0.3 in the visible), assuming a reference height in an aerosol-free area, such as the upper troposphere. In these conditions, the Klett inversion technique [35] is employed to extract the vertical profile of the aerosol backscatter coefficient (b_aer) at the respective wavelengths (355, 532, and 1064 nm) during daytime, assuming a constant lidar ratio profile. The resulting average uncertainty of the retrieved b_aer values (incorporating both statistical and systematic errors over a 30–60 min averaging time) in the troposphere is of the order of 20–30% [36]. To address this substantial uncertainty, the Raman N₂ lidar technique is applied during nighttime, when the atmospheric background is low, given the relatively weak nature of Raman signals.

In the case of the Raman technique, the measurements of the elastic (355 and 532 nm) and the N₂ Raman backscatter signals (387 and 607 nm) allow for the retrieval of the vertical profiles of the aerosol extinction (a_aer) and backscatter coefficients independently of each other [37,38,39], and, thus, of the lidar ratio (LR = a_aer/b_aer) at 355 and 532 nm. The uncertainties associated with the retrieved b_aer and a_aer vertical profiles are of the order of 10–20% and 10–15%, respectively [39,40]. Therefore, the corresponding LR uncertainties range from 14 to 25%, while for the AE related to backscatter (AE_b) and extinction (AE_e), they range from 20 to 35% and 20 to 30%, respectively.

Additionally, the mobile elastic depolarization lidar system AIAS of NTUA was used to provide the vertical δ_aer in the troposphere, and to evaluate the sphericity of the examined aerosols [41,42]. AIAS operates on a pulsed Nd:YAG laser, emitting at 532 nm, with an energy of 98 mJ pulse⁻¹ at a 10 Hz repetition frequency. The laser beam achieves a vertical polarization of (>99%) through a zero-order λ/2 waveplate. Subsequently, it is expanded using a Galilean telescope (×4) before being emitted into the atmosphere, exhibiting a divergence of less than 0.5 mrad. A 200 mm Dall–Kirkham Cassegrainian telescope (focal length f = 1000 mm) collects the elastically backscattered light at 2 polarization planes: parallel and perpendicular, with respect to the emitted polarization plane. A secondary mirror directs the collected light to the collimating lenses, dichroic beam splitters, polarizing cubes, doublets, and interference filters (IFF) placed in front of the PMTs. The polarizing cube beam splitter, attached to a rotatable base, is employed for the separation of the co-polarized and de-polarized light at 532 nm. AIAS uses the ±45° (±0.2° uncertainty) calibration technique, as described in [41,42]. The retrieved δ_aer values exhibit a relative uncertainty of the order of 15%.

2.2. The Flexible Particle Air Mass Dispersion Model (FLEXPART)

To identify the aerosol sources of the case studies presented in Section 3, we used the Lagrangian particle dispersion model FLEXPART, version 10.4 [43,44]. The model was driven by ERA5 [45] assimilated meteorological analyses from the European Centre for Medium-Range Weather Forecasts (ECMWF) with 137 vertical layers, a horizontal resolution of 0.5 × 0.5, and hourly temporal resolution. FLEXPART releases computational particles at 0–5000 m heights from the receptor (NTUA) that are tracked backward in time using FLEXPART’s “retroplume” mode. Simulations extended over 30 days backward in time, sufficient to include most BC and dust emissions arriving at the station given a typical aerosol lifetime of 1 week [14]. For each observation and height, the source–receptor matrix (SRM) (also known as “footprint emission sensitivity”, or “footprint”) was calculated; this resulted in a modeled concentration at the receptor when coupled with gridded emissions from an emission inventory. The emission sensitivity expresses the probability of any release occurring in each grid-cell to arrive at the receptor. FLEXPART considers gravitational settling for spherical particles of the observed size, and differs from trajectory models due to its ability to simulate dry and wet deposition of gases or aerosols, turbulence, and unresolved mesoscale motions, while it includes a deep convection scheme [43,44].

The source contributions to the modeled receptor BC were derived by combining each gridded emission sector from the ECLIPSEv6 emission inventory [46] for anthropogenic sources (namely, gas flaring, transportation, shipping, industrial combustion, transportation, domestic combustion, and waste management) and GFEDv4 emission inventory for biomass burning [47] with the footprint emission sensitivity. Dust emissions were computed 3-hourly at 0.25 degrees of spatial resolution using FLEXDUST [48] driven by ECMWF operational analysis fields at 0.25 degrees. Dust particles with diameters up to 17.32 micrometers were included in this study.

2.3. The MODIS Instrument

The Moderate Resolution Imaging Spectroradiometer (MODIS) instruments on board Terra and Aqua satellites [49] provided active fire data throughout the study period. The data distribution occurred through the Fire Information for Resource Management System (FIRMS) (https://firms.modaps.eosdis.nasa.gov/map (accessed on 25 January 2024)). The confidence of the fire data was selected to be above 80% in order to provide a high level of assurance regarding the active fires and to ensure that the air masses arriving to Athens overpassed areas that enriched the layers with biomass-burning aerosols, where fire hot spots were detected [50,51].

2.4. The CIMEL Sun Photometer

Sun photometric observations were performed using a CIMEL sun–sky radiometer, part of the Aerosol Robotic Network (AERONET) Global Network (http://aeronet.gsfc.nasa.gov (accessed on 25 January 2023)) [52]. The instrument was located at the Research Center for Atmospheric Physics and Climatology of the Academy of Athens (37.97° N, 23.71° E, 130 m a.s.l). The CIMEL sky–sun photometer is an automated, ground-based radiometer designed to measure the direct solar irradiance and diffuse sky radiance along almucantar and principal solar planes, covering a field of view of 1.20°. The standard measurement of CIMEL involves obtaining sets of direct sun measurements every 15 min and recording sky diffuse almucantar or principal plane data every 30 min. The specific wavelengths used in this instrument’s channels may vary depending on the version, but it always includes filters at 440, 675, 870, 940, and 1020 nm. The data utilized in this study after cloud screening are level 1.5 (AOD, AE, aerosol size distributions, and aerosol microphysical properties) [52]. The aerosol size distributions obtained through inversion correspond to aerosol radii spanning from 0.01 to 15 μm. The anticipated precision of the inversions is within the range of 15–25% for aerosol radii greater than 0.5 μm and 25–100% for radii less than 0.5 μm [52,53]. The uncertainty in AOD is <±0.01 for wavelengths higher than 440 nm [54], or approximately 10% for a nominal AOD of 0.1. Aerosol size distribution values are retrieved by the sky radiance measurements, and the uncertainty is assumed to be <±5% at all four wavelength channels [55,56].

2.5. In Situ Aerosol Measurements

To retrieve the real-time concentration of the equivalent black carbon (EBC), we used the Aethalometer A33 (Magee Scientific, Berkeley, CA, USA), which is a 7-wavelength dual spot instrument that collects aerosol samples at 5 lt min⁻¹ through short conductive tubing on a filter tape, absorbing at 880 nm. The time resolution of the measurements was 1 min, which was then averaged on an hourly basis. Elemental and organic carbon (EC and OC, respectively) concentrations were obtained via a Thermo-optical analyzer (Lab OC-EC Aerosol Analyzer model, Sunset Laboratory Inc., Tigard, OR, USA) using PM_2.5 quartz filter samples, while their total mass was obtained via gravimetric analysis. The time resolution of these measurements was 3 h. The EC measurements’ repeatability and reproducibility relative standard deviations were 15 and 20%, respectively [57]. Both instruments were located at Demokritos Station (DEM), a member of the Global Atmospheric Watch (GAW) network; the Aerosol, Clouds, and Trace Gases Research Infrastructure (ACTRIS); and the PANhellenic infrastructure for Atmospheric Composition and climatE chAnge (PANACEA). The DEM monitoring site belongs to the National Centre of Scientific Research, “Demokritos”, which is situated 7 km to the north from downtown Athens, in a pine forest (37.995° N, 23.816° E, at 270 m a.s.l.). It is representative of the urban background atmospheric aerosol concentration in the Athens metropolitan area [58].

Furthermore, aerosol samples were also collected at NTUA (School of Mining and Metallurgical Engineering), 14 m a.g.l. The sampling location was fully exposed to the wind and was free of obstacles. Between 21 and 23 May, ten PM₁₀ and PM_2.5 samples were collected over 3 to 5 h on PTFE membranes following the sampling procedure described in [59,60]. The PM₁₀ and PM_2.5 concentrations were retrieved using the methodology detailed in [60]. Additionally, the concentrations of Cl⁻, NO₃⁻, PO₄⁻³, SO₄²⁻, Na⁺, NH₄⁺, K⁺, Ca²⁺, and Mg²⁺ were determined using ion chromatography, as described in [60].

2.6. Retrieval of the Profiles of the Microphysical Aerosol Properties

The aerosols’ microphysical properties within the dust and/or biomass-burning layers were retrieved using the regularization inversion technique [30,61], using the vertical profiles of a_aer (at 355–532 nm) and b_aer (at 355, 532 and 1064 nm) obtained from the Raman backscattered lidar signals, as well as the depolarization lidar profiles at 532 nm, as input. The inverted aerosol microphysical properties comprised the effective radius (r_eff), the total number (N), the surface area (S) and volume (V), as well as the real (Re[m]) and imaginary (Im[m]) parts of the particle refractive index (m) and the single-scattering albedo (SSA) within the different aerosol layers. In our method, we did not consider the spectral dependence of the m or the chemical composition of the aerosol particles. Therefore, the obtained m values represent the average values considering the size and spectral range (355–1064 nm). The particles were considered as spheres (δ_aer < 10%) and spheroids (δ_aer > 10%). We should also note that for the fine mode, the optical properties of the spheres and spheroids were very similar, and significant differences were found only for the coarse mode. The uncertainty of the Re[m] is of the order of ±0.05, while for Im[m], it is of the order of ±50%. The corresponding uncertainties of the r_eff, V, and S are below 20%, while the uncertainty of N can be up to 50% [61].

2.7. Aerosol Mass Concentration Lidar Retrievals

To estimate the aerosol mass concentration profiles (for dust and non-dust components) [62], we used the polarization lidar photometer networking (POLIPHON) algorithm as detailed in [63]. The method relies on the combined data from depolarization lidar measurements and sun photometry observations. The fundamental concept behind this approach is to utilize depolarization lidar data to distinguish the contributions of dust and non-dust aerosol types in the total aerosol backscatter coefficient profile. The sun–sky photometer provides spectrally resolved, column-integrated particle extinction values (AOD) and allows for the retrieval of fine and coarse mode AODs [56,64], as well as the microphysical aerosol properties, including particle volume (v) and surface-area concentrations for the fine- and coarse-mode fractions. Subsequently, the mass concentrations of the dust components are calculated using their well-established properties, including δ_aer, LR, mass density, and v/AOD. These properties are estimated based on the sun photometric observations.

In this study, we used a coarse-mode particle density of ρ_c = 2.6 g cm⁻³ (desert and mixed dust) [63,65,66,67,68,69] and a fine-mode particle density of ρ_f = 1.35 g cm⁻³ [70,71,72,73]. The latter measurements were specifically employed to obtain the intensive aerosol properties. Consequently, satisfactory results can be obtained even when the lidar and sun photometer measurements are not precisely collocated and simultaneously measured. This is particularly applicable to Sahara dust, given that the v/AOD has been observed to remain relatively stable during specific dust events [63]. The uncertainties associated with the retrieved particle mass concentration primarily include two components [62,74,75]: (i) uncertainties in the retrieved backscattering coefficient and depolarization ratio, mainly influenced by the signal-to-noise ratio of the lidar backscattering signals, and (ii) uncertainties attributed to the input parameters of the POLIPHON method [76]. The uncertainties of the backscatter and extinction coefficient as well as the depolarization ratio at 532 nm arose from the discussions in Section 2.1. A reasonable uncertainty of 20% was assumed for the smoke and dust mass density [63,77]. The conversion factor (v/AOD, volume concentration to AOD ratio), which is also required as input, may vary up to 10% and 20% for dust and smoke, respectively [78]. Taking into consideration all of the above, by applying the law of propagation, we found a comprehensive uncertainty of ~36–40% in the particle mass concentration profiles.

3. Case Study: 21–26 May 2014

In the following sub-sections, we present a 6-day period of mixed aerosols (biomass burning and polluted dust) observed over Athens on 21–26 May 2014 at the 1.3–4.3 km height range above sea level (a.s.l.). The atmospheric structure during this period was characterized by the formation of a low-pressure system over the East Balkans and the Black Sea on 14–17 May 2014. This type of circulation favors the transport of air masses from Kazakhstan towards Greece. This pattern was replaced by a prevailing high-pressure system over the central and eastern Mediterranean during the following days (18–25 May 2014). This system enabled the transport of Saharan dust towards Greece. The combination of the above weather types was responsible for the transport of the biomass and Saharan dust aerosols that were detected over Athens during the measuring period (Figures S1 and S2).

The remote sensing and in situ ground instruments were operating continuously during this period, except for the night of 25 May due to extensive cloud cover conditions over the measuring site.

Figure 1 presents the spatio-temporal evolution of the range-corrected lidar signal (in arbitrary units-AU) obtained using EOLE at 1064 nm from 0.3 to 6.0 km height a.s.l., between 21 May and 26 May 2014. On 21 May, the first day of the studied event, two dispersed free tropospheric aerosol layers were observed within ~2.0–6.0 km over a convective planetary boundary layer (PBL) extending up to ~1.8 km in height (06:10–14:54 UTC), while later, during the night hours (18:44–21:50 UTC), the aerosols were confined inside a quite shallow PBL from the ground up to ~1.7 km (Figure 1, upper left). On 22 May, a convection lifted aerosols up to ~2.2 km height in the PBL (from 06:30 to 11:30 UTC), while aerosol remnants were found at 2.2–2.9 km height (~15:00 UTC) (Figure 1, upper middle). Later the same day (18:27–22:00 UTC), the PBL collapsed, leaving a residual aerosol layer extending from 0.8–1.2 to 3.0 km a.s.l. aloft over the PBL top-height.

On 23 May, a deep convection was again formed over Athens (07:25 and 12:32 UTC), uplifting aerosols up to a height of ~3.0 km, where they were mixed with a residual aerosol layer observed at the same height (1.0–3.0 km) during the previous night hours (Figure 1, upper right). Later on, during nighttime (19:08–21:45 UTC), a strong residual layer, rich in aerosols, was observed over a very shallow PBL (with a top at ~0.5 km height) leaving a residual aerosol layer at ~1.0–1.2 km height. Above that height, several discrete aerosol layers were observed up to ~3.2 km a.s.l.

On the following day, 24 May, the PBL remained shallow up to a height of ~1.1 km, while between 18:56–21:50 UTC, a residual layer rich in aerosols and decoupled from the PBL was formed around 1 km a.s.l. Above that layer, several aerosol layers were observed between 1.0 and 5.0 km a.s.l., mostly descending. On 25 May, a thermal convection led the PBL top-height up to ~1.8 km a.s.l., around 12:00 UTC. During the night hours (18:42–21:16 UTC), the PBL was confined below ~0.4 km a.s.l. Again, as on the previous day, several descending aerosol layers were observed above the PBL, mainly between 1.8 and 4.2 km (daytime hours), extending up to ~5 km (nighttime hours). Finally, on the last day (26 May), the PBL remained quite shallow (PBL top up to ~1.1 km between 05:51 and 15:08 UTC), above which several aerosol layers were observed up to ~5.0 km height, with a very distinct one descending from ~4.0 km (05:51 UTC) down to ~2.0 km a.s.l. (22:00 UTC).

Figure 2 presents the footprint emission sensitivity for dust, as obtained by FLEXPART for the air masses arriving over Athens between 0.5 and 4.0 km, from 21 to 26 May 2014 (04:00 UTC to 22:00 UTC). The red dots represent active fires retrieved using MODIS (confidence level > 80%), confirming that, in general, the air masses overpassed fire hot spot regions (wildfires/agricultural fires) and the Saharan/Kazakhstan deserts before arriving over the study area.

More specifically, from 21 to 23 May (Figure 2a–c, Figures S3–S6), the “footprint” suggests an increased residence time of air masses along the N.E direction axis (Athens–Black Sea–Caspian Sea–Kazakhstan), including the arid regions around the Aral Sea. The latter is an active salt dust source region due to the desiccation of the lake [79,80,81,82]. Therefore, during this period, the air mass trajectory sampled over Athens was enriched with a mixture of desert dust originating from central Kazakhstan and the arid regions around the Aral Sea, and with BB particles emitted from agricultural fires from South Russia and Kazakhstan, as indicated by the FLEXPART and MODIS data.

Regarding the period from 24 to 26 May (Figure 2d–f, Figures S7–S10), the “footprint” suggests that there were two different aerosol source regions potentially affecting the aerosol concentrations over Athens: the Saharan desert, the dominant one, and the agricultural fires and the Kazakhstan area, with much less sensitivity. Therefore, during that period, the air masses arriving over Athens were a mixture of Saharan and Asian dust and biomass-burning aerosols.

In Figure 3, we present the aerosols’ geometrical and optical properties, obtained using the EOLE and AIAS lidars under cloud-free conditions during the period of 21–26 May 2014. The different color bars denote the tops and the bottoms of the sampled aerosol layers together with the mean values of the AE_b355/532, LR₃₅₅, δ₅₃₂, AE_b532/1064, and LR₅₃₂, as well as their standard deviations.

From the data presented in Figure 3, it can be discerned that the main aerosol layers were found between 1.3 and 4.3 km a.s.l. The retrieved aerosol optical properties showed that the elevated layers (on 21–22 May) were characterized by LRs from 25 to 54 sr (355 nm) (Figure 3c) and 31 to 56 sr (532 nm) (Figure 3f). The δ₅₃₂ values ranged between 5 and 9% (Figure 3d), lower than the typical values [41] usually measured for pure dust. The AE_b355/532 varied from 0.82 to 1.50 (Figure 3b), and the AE_b532/1064 from 0.91 to 1.48 (Figure 3e), which indicated a presence of fine-mode particles.

Particularly on 23 May, high δ₅₃₂ values (19–27%) indicative of the presence of mixed dust particles were observed on that day, which is corroborated by the LR values, which ranged between 33 and 52 sr (355 nm) (Figure 3c) and 41 and 62 sr (532 nm) (Figure 3f). We should also note that the LR₃₅₅/LR₅₃₂ ratio (color ratio) dropped below 1 on that day and ranged between 0.65 to 0.87, which is indicative of fresh biomass-burning air masses [83]. Furthermore, the AE varied from 1 to 2.01, from 1.01 to 1.21 (AE_b355/532) (Figure 3b), and from 0.91 to 1.48 (AE_b532/1064) (Figure 3e). The significantly low values of the color ratio along with the aforementioned AE values are indicative of the presence of mixed biomass-burning and dust aerosols [84,85,86,87], which were found to originate from North Kazakhstan and South Russia, as determined via the FLEXPART simulations presented above (Figure 2).

During the subsequent period (24–26 May), when the air mass trajectories were traced back to the Saharan desert and South Russia, a change in the aerosol optical properties was noted. The LRs increased by almost two times compared to the days before, varying from 84 to 92 sr at 355 nm (Figure 3c) and from 37 to 70 sr at 532 nm (Figure 3c,f), implying the presence of strongly absorbing particles [23]. Concurrently, the δ₅₃₂ exhibited values from 7 to 28% (24–26 May) (Figure 3d) and the AE ranged between −0.36 and 1.68 (AE_b355/532) (Figure 3b) and 0.63 and 2.01 (AE_b532/1064) (Figure 3e). Furthermore, the LR₃₅₅/LR₅₃₂ for these days increased above 1 (1.3–1.8), indicating the presence of aged biomass-burning particles. Specifically, on 26 May, the color ratio values were accompanied by significantly low AE_b355/532 values, which ranged between −0.36 to 0.44, revealing the domination of large dust particles [83,86,88]. These observations corroborate the presence of both polluted dust and smoke aerosols [23,24], as the air mass trajectories originated from the Saharan desert and South Russia.

In Figure 4, we present the temporal variation of the hourly averaged values of OC, EC, and EBC (Figure 4a) accompanied by the daily AOD at visible, with the corresponding contributions from the fine and coarse modes and the AE at 440/870 nm (Figure 4b). The EBC and EC covaried throughout the measurement period (EBC~0.13–1.0 μg m⁻³, EC~0.14–1.48 μg m⁻³). Notably, during the period of 21–23 May, the mass concentrations of EBC ranged between 0.13 and 0.77 μg m⁻³, and those of EC varied from 0.14 to 0.82 μg m⁻³. For the following days (24–26 May), slightly elevated mass concentrations were observed for both EBC and EC, spanning 0.30–1.0 μg m⁻³ and 0.30–1.48 μg m⁻³, respectively. Regarding the OC concentrations, a similar pattern was observed for both periods, but with consistently higher mass concentrations (OC~0.81–4.44 μg m⁻³) compared to the EBC and EC. Specifically, we found that OC varied between 0.81 and 2.46 μg m⁻³ for the period of 21–23 May, and during 24–26 May, an increase in mass concentrations was noted, ranging from 1.34 to 4.4 μg m⁻³.

The differences in mass concentrations observed during these two periods are due to variations in the sources and the trajectories of the air masses prior to their arrival over Athens. As discussed already in Figure 2, FLEXPART “footprint” analysis outlined that, during the first period (21–23 May) the air masses originating from the Kazakhstan area traversed extensive distances, thus resulting in relatively diminished pollutant concentrations upon reaching Athens. In contrast, during the second period (24–26 May), FLEXPART revealed the diverse origin of the air masses, emanating from the Saharan desert, South Russia, and North Kazakhstan. However, the predominance of air masses originating from the Saharan desert may have potentially led to a lower pollutant dispersion and higher mass concentrations. It should be noted that the PBL was significantly shallower in the second period in comparison to the first period (Figure 1). The higher EBC, OC, and EC concentration values observed during the second period could be partly attributed to this fact.

On 23 May, the concentration of PM_2.5 reached the maximum 24 h threshold of 25 μg m⁻³ at 14 m above ground level (Figures S12 and S13), indicating that, at ground level, the concentration of PM_2.5 was even higher. This was due to the addition of the influx of fine particles transported from North Kazakhstan and South Russia in the local aerosol. Furthermore, the PM₁₀ concentration also reached a high value (42 μg m⁻³) due to the presence of coarse particles transported from Kazakhstan Desert [89,90]. Between 21 and 23 May, the observed high concentrations of NO₃⁻ (317–1368 ng m⁻³ at PM_2.5), NH₄⁺ (829–1311 ng m⁻³ at PM_2.5, 1111–1663 ng m⁻³ at PM₁₀), and SO₄²⁻ (3233–5488 ng m⁻³ at PM_2.5, 3798–4204 ng m⁻³ at PM₁₀) (Figure S14) reflected the presence of biomass-burning aerosols near the ground. More specifically, the NH₄⁺ concentrations reflected the presence of biomass-burning and marine aerosols mixed with aerosols originating from the Black Sea [91,92]. Moreover, the concentration of SO₄²⁻ depends upon local and long-range transport of anthropogenically polluted air masses [93,94].

On the other hand, Ca²⁺ and Mg²⁺ ions are present in significant amounts in mineral dust particles [95]. During the study period, a sharp increase in the concentrations of Ca²⁺ (103–502 ng m⁻³ at PM_2.5, 370–1484 ng m⁻³ at PM₁₀) and Mg²⁺ (14–139 ng m⁻³ at PM_2.5, 33–417 ng m⁻³ at PM₁₀) was observed in both fine and coarse aerosol samples. The increase was more pronounced for PM₁₀, marking the presence of coarse particles due to dust transportation over Athens. K⁺ is an element with a mixed origin in the aerosol: it is contained in the aerosol originating from biomass burning, but it is also contained in dust particles [96,97,98,99]. K⁺ concentration levels were also elevated in both PM_2.5 and PM₁₀ samples (53–114 ng m⁻³ at PM_2.5, 48–150 ng m⁻³ at PM₁₀).

Moreover, in Figure 4b, the AOD values at 500 nm show strong variability. The maximum values of the AOD (~0.3) were found on 26 May, which is indicative of the presence of aerosols reducing the air quality over the Athens Basin [100]. The coarse mode contribution to the total AOD reached its maximum value on 23 (57%) and 26 (61%) May due to presence of mixed Asian and Saharan polluted dust [101], respectively, over Athens, with a corresponding decrease in the AE on these days (daily mean values of 0.76 and 0.61, respectively). The fine-mode particles within the atmospheric column accounted for 64–77% on the remaining days, indicating the dominance of smaller particles, while the mean AE increased up to about 1.45.

Additionally, in Figure 5a, we present the corresponding columnar mean size distribution retrieved by sun photometric measurements over Athens for the period of 21–26 May 2014. The bi-modal size distribution shows particles with radii ranging from 0.14 to 0.19 µm and around 1.8 µm). Specifically, on 21–22 May, the fine particles dominated due to the presence of smoke, while on 23–26 May, the coarse mode became more dominant due to the mixing of smoke and desert dust aerosols. On both 23 May and 26 May, the columnar mean size distribution revealed a predominance of larger particles (with effective radii centered around 1.7 µm) in comparison to smaller ones (with effective radii centered around 0.15 µm). When we compare the experimental size distributions obtained using 24 h cascade impactor aerosol surface measurements at the same location under similar conditions (Figure S17), as discussed in [102], we can observe that the same bi-modal behavior is confirmed at the surface observations in this size range, with a 3rd minor peak appearing above 10 µm. In terms of aerodynamic diameter, the submicron modes in all cases were smaller (0.27–0.3 µm) in the retrieved size distribution compared to the observed size distributions at the surface (0.4 and 0.37 µm for Sahara and background cases, respectively [102]). This can be explained by the difference in aerodynamic and optical diameters used in the two descriptions. This difference was not observed in the coarse mode, where, in all cases, the peak was located between 3.4 and 3.8 µm. This may have been caused by the stronger variability in the coarse particle density between the surface and elevated layers and could be a sensitivity parameter to be tested when elaborating the retrieval algorithms. The chemical compositions observed for Saharan and background aerosol were very similar to that observed at the same location in previous earlier studies [103]. It is confirmed that the distribution of ions like SO₄, existing almost exclusively in the PM_2.5 fraction (Figure S15, Table S1) or Ca²⁺ in the coarse fraction (Figure S16, Table S2), was the same in the current study.

The daily mass-specific extinction coefficient (k_ext) for the coarse and fine modes, as derived from CIMEL sun photometer during the period of 21–26 May 2014, is presented in Figure 5b. The mass-specific extinction coefficient holds a coarse-mode particle density of ρ_c = 2.6 g cm⁻³ (desert and mixed dust) [63,65,66,67,68] and a fine-mode particle density of ρ_f = 1.35 g cm⁻³ [70,71,72,73]. We observed that the k_ext values of the coarse particles ranged between 0.32 and 0.58 m² g⁻¹. Similar values have been reported for dust particles [63,65,104,105], leading to the conclusion that the k_ext values are independent of the distance of the observation from the source [63]. Regarding the fine mode, the k_ext values ranged between 2.58 and 4.79 m² g⁻¹, indicative of the presence of biomass-burning smoke and/or urban haze [65,76,106]. The dispersion of the k_ext values for a specific aerosol type can offer insight into the range of uncertainty in the particle mass concentration retrievals when incorporating values from existing literature in the data analysis. The uncertainties related to the retrieval of the k_ext values and, therefore, to the particle mass concentration, mainly arise from uncertainties related to the conversion of the backscatter into extinction coefficients caused by the lidar ratio estimation, as well as from the used v/AOD ratio (conversion factor), as discussed in Section 2.7. More details about the errors in the retrieval of mass concentrations can be found in [76].

3.1. Case of 23 May 2014: Aged BB and Dust (Russian Forest Fires and Kazakhstan Dust Aerosols)

The retrieved profiles of b_aer (Figure 6a) and δ₅₃₂ (Figure 6b) on 23 May (20:30–21:45 UTC) showed the presence of four elevated layers, characterized by high aerosol backscatter coefficients (Figure S11) and values of δ₅₃₂ ranging from 19–27%, indicating the presence of mixed biomass-burning and dust aerosols [33,62,87,107]. In order to identify the various aerosol layers, we used the gradient method [108] by studying the slope of the backscatter coefficient. The identified layers were found to be: layer 1 (1.32–1.62 km), layer 2 (2.22–2.76 km), layer 3 (3.06–3.48 km), and layer 4 (3.60–3.90 km). Additionally, the inversion algorithm was applied to the mean-layer-derived values of the aerosol optical properties (3b + 2a + δ) to estimate the corresponding mean values of the microphysical properties (r_eff, m, SSA).

The r_eff retrievals (Figure 6c) indicate the vertical homogenous conditions of the mixture of fine and coarse modes at the first, third, and fourth aerosol layers (r_eff 0.22 ± 0.04 μm). However, at the second aerosol layer at 2.46 km, the value of r_eff was significantly lower (r_eff 0.14 ± 0.03 μm), but with the highest observed number density of 1363 ± 333 cm⁻³ (not shown), showing a significant contribution of fine mode particles due to the mixture of aged, biomass-burning aerosols with dust. Furthermore, a decrease in the number density for the third and fourth layers (70 ± 21 cm⁻³ and 140 ± 42 cm⁻³, respectively), indicate the mixing of dust with biomass-burning particles [24].

The real part of the refractive index (Figure 6d) was found to be 1.61 ± 0.05 for all layers, indicating a strong mixing of dust with urban-like sulfate and organic carbon aerosols over Athens [23,109,110]. The imaginary parts of the refractive index were found to be 0.003 ± 0.002, 0.004 ± 0.002, and 0.002 ± 0.001 for layers 1, 2, and 3, respectively, while for layer 4, it increased up to 0.005 ± 0.003, indicating the presence of aerosols mixed with slightly absorbing dust [10,111,112,113,114]. The mean SSA values were found to be 0.951 ± 0.195 for all aerosol layers (Figure 6e), confirming the presence of slightly absorbing particles. This is in agreement with a previous observation of Russian fires in Athens in 2010 [115], with elevated SSA values at 532 nm close to 0.91 during the period of the fires. The retrieved microphysical data can be further used as input to the ISORROPIA model to obtain the chemical composition of the aerosols aloft [32,33].

In Figure 7, we present the b_aer for the total, fine, and coarse modes, as well as δ₅₃₂, as derived by the AIAS lidar system, together with a_aer and the mass concentration (fine and coarse) on 23 May (14:30–15:30 UTC), using the POLIPHON method discussed in Section 2.7. The vertical profile of total b_aer (Figure 7a) indicates moderate backscattering up to 3.5 km. Above a height of 1.5 km, the smoke and dust particles contributed almost equally to the backscattering coefficient. The δ₅₃₂ (Figure 7b) ranging between 16–27% is indicative of a mixture of smoke and dust particles [23,84,86] coming from North Kazakhstan and South Russia, as previously discussed (Figure 2). However, the relevant fine mode (smoke) extinction coefficients (Figure 7c) are rather higher than the coarse (dust) ones, since the smoke lidar ratio was larger than the desert dust one. More specifically, lidar ratios of 75 and 55s were employed to convert backscatter into extinction coefficients, typical for biomass-burning particles (fine mode) and desert dust (coarse mode), respectively [116,117,118,119,120].

Furthermore, to retrieve the aerosol mass concentrations (Figure 7d), we used densities of ρ_f = 1.35 g cm⁻³ and ρ_c = 2.6 g cm⁻³ for fine (smoke) and coarse (dust) particles [70,71,72,73]. Afterwards, we calculated the mean volume concentration to AOD ratio values equal to v_c/AOD_c = 0.79 × 10⁻⁶ m (coarse-mode) and v_f/AOD_f = 0.21 × 10⁻⁶ m (fine-mode). The mean relevant mass-specific extinction coefficients obtained at the nearest temporal resolution were k_ext,c = 0.50 m² g⁻¹ and k_ext,f = 3.05 m² g⁻¹ for coarse-mode and fine-mode particles, respectively, for that day. Thus, a mean smoke mass concentration of 10 μg m⁻³ was estimated, while for dust, the mass concentration ranged between 10 ± 4 and 160 ± 58 μg m⁻³ (for a mean AOD₅₃₂ of 0.17). Below a height of 1.5 km, the mean value of smoke mass concentration was found to be approximately equal to 16 ± 6 μg m⁻³, very close to the PM_2.5 values (24.5 μg m⁻³) measured at ground level, which were slightly higher than the World Health Organization (WHO) air quality levels (24 h average less than 15 μg m⁻³).

On the other hand, the lower-atmospheric mean dust mass concentration values, found to be about 150 ± 54 μg m⁻³ (around 1.5 km height), were almost three times higher than the PM₁₀ (42 μg m⁻³) mean concentration at the surface, despite all the different sources of uncertainty for such a comparison.

3.2. Case 26 May 2014: Aged BB and Dust (Unusual Sahara and Kazakhstan Dust Aerosols)

In Figure 8a,b, we present the aerosol optical profiles (3b + 2a + δ) from 26 May, averaged within 19:00–21:45 UTC. Four different layers were identified: layer 1 (1.68–2.04 km), layer 2 (2.28–2.58 km), layer 3 (3.12–3.66 km), and layer 4 (3.90–4.32 km).

The retrieved aerosol properties (Figure 8a,b) show that the four elevated layers were mostly composed of mixed biomass-burning and dust aerosols, since δ₅₃₂ values were found to take from very low to very high values (7–28%).

The r_eff estimations at the first two layers (centered around1.8 to 2.6 km altitude) showed a significant dominance of fine-mode particles (r_eff 0.15 ± 0.03 μm) (Figure 8c), accompanied by a high number density of 1741 ± 520 cm⁻³. However, in the two upper layers (above 3 km), the r_eff ranged from 0.22 to 0.27 μm, indicating the predominance of dust particles accompanied by a smaller number density (455 ± 137 cm⁻³), as estimated for both layers.

Furthermore, in Figure 8d, we present the retrieved refractive indices, spanning a range from 1.55 ± 0.05 to 1.61 ± 0.05 for the real part and between 0.006 ± 0.003 to 0.008 ± 0.004 for the imaginary part. Despite the relatively low Im[m] values, they still highlight the presence of aerosols with absorption capabilities. The SSA mean values were consistently found to be 0.96 ± 0.19 for all four layers (Figure 6e), again indicating the presence of low-absorbing particles, as corroborated by previous findings [20].

In Figure 9, we illustrate four distinct aerosol layers characterized by low-to-moderate backscattering (0.3–2.0 Mm⁻¹ sr⁻¹), reaching an altitude of about 4.5 km a.s.l. (Figure 9a). The first two layers (1.9–2.4 km) exhibited 14–17% depolarization ratios (Figure 9b), typical values associated with a mixture of fine and coarse particles, with the fine mode predominating. The two uppermost layers (3.4–4.1 km) showed higher depolarization ratios (20–28%), indicative of non-spherical dust particles [20,86].

The fine and coarse mode mean mass specific extinction coefficients were k_ext,c = 0.57 m² g⁻¹ and k_ext,f = 3.10 m² g⁻¹, with mean volume concentrations to AOD values equal to v_c/AOD_c = 0.67 × 10⁻⁶ m (coarse-mode) and v_f/AOD_f = 0.20 × 10⁻⁶ m (fine-mode). As in the previous case of 23 May, lidar ratios of 75 sr and 55 sr were used for the conversion of the aerosol backscatter into the particle extinction coefficients (Figure 9c) for the biomass-burning (fine mode) and desert dust (coarse mode) particles, respectively. Thus, the dust extinction coefficients were of the order of 30 Mm⁻¹ and 50 Mm⁻¹ for fine particles.

The elevated dust mass concentrations ranged from 40 ± 16 to 120 ± 43 μg m⁻³ (Figure 9d), with increasing values up to 3.5 km. Regarding the smoke mass concentration, a mean value of about 20 ± 8 μg m⁻³ was estimated, with the maximum 30 ± 12 μg m⁻³ taken below 1.5 km of height. In this case, aged biomass-burning aerosols originating from wildfires in South Russia considerably contributed to the fine mode particle mass. The mean total AOD₅₃₂ derived by the AIAS lidar was found to be 0.25.

3.3. Intercomparison of Aerosol Columnar Retrievals (Lidar and Sun Photometer)

In Figure 10, we discuss the similarity of the aerosol properties derived by the lidar with the ones obtained by CIMEL. Specifically, we present the temporal evolution of the refractive indices Re[m], Im[m], r_eff, and SSA (532 and 440 nm), as well as the columnar daily mean AOD (340, 500, 1020 nm) values, using data from both instruments (EOLE lidar and CIMEL sun photometer) for the period of 21–26 May 2014.

More specifically, Figure 10a presents the real part of m obtained by inverting EOLE lidar signals. The values ranged from 1.44 ± 0.05 to 1.61 ± 0.05 and revealed the presence of mixed dust with aged biomass-burning aerosols [83,86,121]. The corresponding values obtained using the sun photometer at various wavelengths (440, 675, 870, and 1020 nm) varied between 1.45 ± 0.07 and 1.50 ± 0.07 and were in accordance with the lidar inversion estimates. The imaginary part (Figure 10b) varied between 0.004 ± 0.002 and 0.008 ± 0.004, indicating the presence of low-absorbing particles. However, the integrated values of the imaginary part, as obtained by the sun photometer, ranged from 0.0005 to 0.002, an order of magnitude lower compared to the ones derived by the lidar. We should note that the uncertainty of Im[m] depends on its variation due to the wavelength dependence, and its uncertainty increases when non-spherical particles are considered in the retrieval [23]. Thus, in our case, a straightforward evaluation of the vertical variation of Im[m] was not possible due to the significant uncertainty [23].

The lidar-derived SSA ranged from 0.94 ± 0.19 to 0.97 ± 0.20 (Figure 10c), indicative of a mixture of dust and biomass-burning particles [20,24,86]. The corresponding SSA sun photometric retrievals (at 440 nm) showed a mean value of about 0.99 ± 0.05, in a good agreement with the lidar estimations.

In Figure 10d, we present the AOD as obtained using the sun photometer and the Raman lidar. We can observe that the AOD values (from 340 to 1020 nm) increased from about 0.03 to 0.50 (340 nm) and from 0.01 to 0.34 (500 nm), while those derived by the lidar ranged from 0.09–0.43 (355 nm) to 0.07–0.25 (532 nm). Despite the relatively good agreement between the two retrievals, it seems that there were some AOD discrepancies, probably arising from various sources, like (a) measurements that were not temporally synchronous, as -sun photometric data were obtained during daytime and Raman lidar measurements at nighttime; (b) different operation principle between the two instruments—the sun photometer measured the direct solar irradiance in the entire atmospheric column to obtain the AOD, while the layer-derived AOD from lidar was estimated by calculating the integral of the extinction coefficient from heights of ~0.5 to 5.0 km a.s.l. (in our case).

4. Conclusions

This study presents an unusual case of mixed biomass-burning and polluted dust aerosols transported from Kazakhstan and Saharan deserts over Athens, Greece (21 to 26 May 2014). To study this event, we used a synergy of remote sensing instrumentation (multi-wavelength elastic-Raman-depolarization lidar, CIMEL sun photometer) and in situ aerosol sampling, as well as models (microphysical inversion schemes), to retrieve a complete set of the aerosols’ optical, microphysical, and chemical properties. The FLEXPART dispersion model analysis highlighted two cases of mixtures: (a) one of biomass-burning and dust aerosols originating predominantly from the South Russia–North and central Kazakhstan and partially from the Saharan desert region (23 May); and (b) one of mixed particles coming from the Saharan desert, with a minor contribution from the southern Russia and northern Kazakhstan areas (26 May). From the lidar observations, we found LR values ranging from 25 to 89 sr (355 nm) and 35 to 70 sr (532 nm), while the δ₅₃₂ values ranged from 7 to 28%. Moreover, the observed AE_b355/532 and AE_b532/1064 values were found to be −0.36–2.01 and 0.63–2.01, respectively. The significant spread of the observed optical properties indicates the mixing of biomass-burning with mineral dust particles, influenced also by the transport time and the nature of the aerosol sources.

The EBC, EC, and OC concentrations covaried throughout the study, with OC concentrations ranging between ~0.81–4.44 μg m⁻³, influenced by the air mass origin. The analysis of water-soluble ionic species of PM_2.5 and PM₁₀ showed elevated concentrations of NO₃⁻, NH₄⁺, SO₄²⁻, Mg²⁺, and Ca²⁺ on 21 and 23 May, indicating predominantly anthropogenic and coarse particles.

The columnar mean size distribution retrieved using sun photometric measurements over Athens revealed that, on 21–22 May, the fine smoke particles dominated (~0.15 μm), while on 23–26 May, the coarse mode became more pronounced (~1.7 μm). The POLIPHON algorithm was used to estimate the vertical profile of the mass concentration for the dust and non-dust components on 23 and 26 May.

Regarding the layer-derived aerosol microphysical properties, we found that the refractive indices ranged from 1.39 (±0.05) + 0.002 (±0.001)i to 1.63 (±0.05) + 0.008 (±0.004)i, with a mean SSA value (532 nm) of approximately 0.96 ± 0.19, indicating slightly absorbing particles.

A comparison of the retrieved aerosol properties with sun photometric data demonstrated good agreement regarding SSA and Re[m]. Although evaluating the variation of Im[m] is challenging due to higher uncertainties, the AOD and AE (440/870) showed very good agreement between the two retrieval methods.

Our study suggests the need for further synergy between in situ aerosol chemical data and multi-wavelength Raman-depolarization lidar measurements to refine the retrieval procedure of unusual mixtures and minimize uncertainties.